High performance computing is helping to deliver exhilarating research projects in Universities. But without investment in the right software and people skills, inefficient HPC will not deliver the research and prestige Universities may expect, explains David Lecomber, CEO and founder of Allinea Software.
World-class research projects are being delivered thanks to High Performance Computing (HPC) systems. Scientific researchers, engineers and academic institutions are using supercomputers to push the boundaries of knowledge ever wider, helping to solve mankind’s super-problems. They can simulate the formation of the first galaxies, predict weather events and work on finding cures for cancer.
There’s no denying that HPC is a brilliant addition to higher education and beyond – but mere ownership of an HPC system does not necessarily lead to prestigious research results or world-changing outputs. Without the investment in tools and training for effective software development and optimization, HPC facilities will not be running at their full potential.
It is vital that when making the significant investment in purchasing HPC systems, that those responsible for the procurement look closely and carefully at ways of extracting the most innovation and performance from the scientific software that is developed and/or deployed – because it is the code that has the potential to really boost results. With researchers so often working to deadlines to produce results for academic conferences, funding bodies and their own career path, it is essential HPC-powered research projects stay on track and are not derailed by failing to use HPC systems to their full capability. Issues in software development, such as poor performance and bugs can hinder progress. It is only with the right software tools – and by training the next generation of scientific software developers in best practices that HPC will really enable projects and simulations to fly, along with the consequent enhancement of academic reputations and the prestige that ground breaking research and simulations bring to our Universities.
It is vital that when making the significant investment in purchasing HPC systems, that those responsible for the procurement look closely and carefully at ways of extracting the most innovation and performance from the scientific software that is developed and/or deployed – because it is the code that has the potential to really boost results
At a recent event, I was talking to an academic who, as a result of analysing their code with newly installed HPC software tools, found a 25% upspeed from their supercomputer in a single afternoon, such is the effect that paying attention to software can have on HPC. Profiling tools and debuggers can dramatically speed up research time by running applications faster. These accessible and easy-to-use tools untap potential without researchers having to learn too much about the detail of their system.
If you are not an expert and the system crashes, valuable time can be taken up figuring out the problem – or, even more frustratingly, it might not be figured out at all. Companies such as Allinea provide software to debug complex projects and highlight performance issues ripe for code optimisation using a profiler. This can analyse how much time, memory or other resources are being used at different points, helping researchers boost the performance of their code by understanding how it is performing and making the right changes at the right time.
At Cambridge University, for instance, research group, the COSMOS consortium, Europe’s leading group of cosmology investigators brought together by Professor Stephen Hawking, has deployed Allinea’s tools, bringing profiling and debugging power to COSMOS_IX, one of the world’s largest shared-memory supercomputers. The tools are very easy to use which is a significant benefit for scientists who need to harness parallel programming, but focus on their science.
The HemeLB research group at University College London (UCL) has also recognised the importance of this type of debugging and optimising software. The group has developed software that applies computational fluid dynamics to model blood flow around cerebral vessels and simulate pressure at points of weakness such as aneurysms. The group collaborated with Allinea to help them address the challenges of application development at scale and resolved an intractable problem on one of the UK’s largest supercomputers, called ARCHER.
A multi-million investment by a University on supercomputer hardware could be compromised if an investment in the software and people is not considered too
Cambridge University and UCL are not the only academic institutions to recognise the huge benefits of this type of software. Scientists at Universities such as Oxford, Princeton, Notre Dame, Edinburgh, Nebraska and Luxembourg are using Allinea Software to enhance their research success.
There is also a growing movement in academic research towards employing specialists whose sole responsibility is to concentrate on HPC software, transforming the millions of pounds worth of investment and producing results that really sing. The UK Community of Research Software Engineers, for instance, has been formed to represent these highly skilled people and raise awareness of their fundamental role in research. A Research Software Engineer essentially works with researchers to gain an understanding of the problems they face, and then develops, maintains and extends software to provide the answers.
In conclusion, a multi-million investment by a University on supercomputer hardware could be compromised if an investment in the software and people is not considered too. Software unlocks the hardware’s true value, unleashing all the benefits that HPC brings; transforming performance, delivering electrifying results quickly and smoothly in ground breaking, reputation-enhancing projects.